170-1354-1-PB

http://jtlu.org. 4 . 1 [Spring 2011] pp. 2544 doi: 10.5198/jtlu.v4i1.170

New urbanism or metropolitan-level centralization?A comparison of the inuences of metropolitan-level and neighborhood-level urban form characteristics ontravel behavior

Petter NssAalborg University, Denmark a

Abstract: Based on a study in the Copenhagen Metropolitan Area, this paper compares the inuences of macro-level and micro-level urbanform characteristics on the respondents traveling distance by car on weekdays. e Copenhagen study shows that metropolitan-scale urban-structural variables generally exert stronger inuences than neighborhood-scale built-environment characteristics on the amount of car travel.In particular, the location of the residence relative to the main city center of the metropolitan region shows a strong eect. Some local scalevariables oen described as inuential in the literature, such as neighborhood street pattern, show no signicant eect on car travel whenprovisions are made to control for the location of the dwelling relative to the city center.

Keywords: Residential location; travel; regional accessibility; centrality; neighborhood characteristics; rationales

1 Introduction

In the United States, research into relationships between landuse and transport during recent years has been largely focusedon the inuence of local-scale urban structural conditions ontravel behavior, comparing traditional suburban residential ar-eas with areas developed according to the so-called New Ur-banism or Transit Oriented Development principles (e.g.Cervero 1989; Krizek 2003). Oen, studies comparing thetravel behavior of residents living in dierent kinds of built en-vironment do not take the location of the investigated neigh-borhoods into consideration. For example, among 38 researchstudies reviewed in a recentAmerican articleCao et al. (2009),only six included variables indicating the locationof theneigh-borhood relative to the city center, and one of these stud-ies was actually European. According to Boarnet and Crane(2001, 49),a relatively limited geographical scale (not muchmore than a census tract) was, when their bookwas published,typical of virtually all recent American empirical research intorelationships between built environment characteristics andtravel.

In a European context, research into relationships betweenland use and travel has focused much more strongly on thelocation of the residence relative to the main metropolitan

[email protected]

center and sub-centers within the metropolitan-scale spatialstructure. Studies in a number of cities in dierent Europeanas well as Asian and South American countries have shownthat residents living close to the city center travel less than theirouter-area counterparts and carry out a higher proportion oftheir travel by bicycle or on foot (e.g. Mogridge 1985; Nss2006b; Nss et al. 1995; Zegras 2010).

Based on a study in the Copenhagen Metropolitan Area,this paper compares the inuences of macro-level and micro-level urban form characteristics on respondents traveling dis-tance by car on weekdays. e main results of the Copen-hagen Metropolitan Area study have been published else-where Nss (2005, 2006a,b, 2009b) and will therefore onlybe presented briey here. e same applies to the theo-retical background and the research methods used. esehave been described in detail in the above-mentioned pub-lications and in a separate paper in which the CopenhagenMetropolitan Area study is used as a reference case for a dis-cussion of the ontological, epistemological and methodologi-cal basis of research into relationships between residential lo-cation and travel (Nss 2004). e present paper concen-trates on a comparison of the eects ofmetropolitan-scale and

is also includes the issue of residential self-selection, which has beenexamined in detail in Nss (2009a) and hence will not be a focus of thepresent paper.

Copyright 2011 Petter Nss.Licensed under the Creative Commons Attribution NonCommercial License 3.0.

.

neighborhood-scale urban form characteristics on the amountof car travel, and an explanation of why the former variablesturn out to be more inuential.

In the next section, the theoretical background of the studyis presented, followed by a section about the case urban region(Copenhagen Metropolitan Area) and the research methods.e empirical results are presented in sections 47.

2 Theoretical background

e so-called activity-based approach (Fox 1995; Jones 1990;Vilhelmson 1999) is a useful conceptual framework for ourstudy. According to this approach, nearly all travel activity isderived from the need orwish to carry out other, stationary ac-tivities. Activities are carried out to fulll physiological needs(eating, sleeping), institutional needs (work, education), per-sonal obligations (childcare, shopping), and personal prefer-ences (leisure activities) (Vilhelmson 1999, 178). In recentyears, this view has been challenged by theorists who considertravel in late modern society to be increasingly a purpose in it-self, rather than an instrument to move from one place to an-other (Steg et al. 2001; Urry 2000). is may be true to someextent for vacation and leisure trips, but the activity-based ap-proach remains, inmyopinion, a useful tool for understandingand analyzing daily travel behavior.

Hgerstrand (1970) distinguishes between three types oftime-geographical restrictions on human activities: capabil-ity constraints, coupling constraints, and authority or steer-ing constraints. Together, the dierent types of restrictionsconstitute a signicant limitation on peoples use of time andon the spatial distribution of their activities, in particular forworkforce participants and pupils on workdays and school-days. Hence, the scope for free activities on weekdays farfrom home is limited, in particular for those who do not havea private motor vehicle at their disposal. erefore, there willbe distance decay in the attractiveness of facilities (Maddi-son et al. 1996). e impact of a remote residential locationin terms of longer traveling distances is therefore likely to becounteracted to some extent by a lower frequency of trips, atleast for non-compulsory activities.

Based on Vilhelmson (1999, 181), trips can be classiedinto dierent categories, depending on how xed or exiblethey are in time and space. Bounded trips are those under-taken in order to reach activities for which both the time andgeographical location are xed and cannot freely be deviatedfrom, e.g. journeys to work or school. According to Vilhelm-son, the majority of trips on weekdays belong to this category.Non-bounded trips are those where the time of the activity

is exible and the location may vary. A heterogeneous inter-mediary group includes trips where the time of the activity ismore or less xed but the location may vary, and trips wherethe location is more or less xed but the time may vary. eextent of space-time xity varies substantially between indi-viduals. For example, although the journey to work has a highdegree of xity formost workforce participants, someworkers(e.g. service mechanics and builders) work at dierent places,and the duration of work at the same location may also vary.

For some facility types, we almost always choose the closestfacility, because the various facilities aremore or less equal (e.g.post oces) or have regulated catchment areas (e.g. local gov-ernment oces). But for other facilities, qualitative or sym-bolic dierences within each facility category may cause peo-ple to travel beyond thenearest facility to amore attractive onefarther away. For cinemas and a number of other recreationalfacilities, many types of shops, and not the least workplaces,a number of features other than proximity are also importantwhen choosing among facilities.

Despite decentralizing trends, most cities still have a higherconcentration of workplaces, retail businesses, public agen-cies, cultural events, and leisure facilities in the historical ur-ban center and its immediate surroundings than in the periph-eral parts of the urban area. For residents in the inner and cen-tral parts of the city, the distances to this concentration of fa-cilities will be short. Downtown is usually also close to the ge-ographical center of gravity of theworkplaces and service facil-ities that are not themselves located to the city center (Nielsen2002). erefore, the average distance to the peripheral work-places and facilities will also be shorter for those living close tothe city center. Local-area densities are usually also higher inthe inner parts of cities than in the peripheral suburbs. Witha higher density of residences or workplaces in the local area,the population base for various types of local service facilitieswill increase. Hence, the average distance from residences tolocal services will also be shorter. Inner-city residents couldthus be expected, on average, to make shorter daily trips thantheir outer-area counterparts to both local and more special-ized facilities, and a high proportion of destinations might beeasily reached on foot or by bicycle.

A large number of empirical studies conducted during thelast couple of decades have concluded that the amount oftravel and the proportion travel by car are higher amongsuburbanites than among inner-city residents. is relation-ship holds true when controlling for demographic and so-cioeconomic variables, and also in the cases where controlhas been made for transport attitudes or residential prefer-ences (Fourchier 1998; Mogridge 1985; Newman and Ken-

New urbanism or metropolitan-level centralization?

worthy 1989; Nielsen 2002; Nss 2006b, 2010; Nss andJensen 2004; Nss et al. 1995; Schwanen et al. 2001; SteadandMarshall 2001; Zegras 2010).

Local-scale urban design principlessuch as street pattern,availability of sidewalks and bicycle paths, etc.as well as aes-thetic neighborhood qualities can inuence the attractivenessof dierent travel modes and can for some travel purposes alsoaect trip destinations. As mentioned above, interest in suchcharacteristics has been at the core of American studies of theinuence of the built environment on travel behavior. Forexample, in their inuential book Travel by Design, Boarnetand Crane (2001, 37) mention the following six urban fea-tures as urban form and land use measures that might inu-ence travel behavior: density (residential or employment, ormore complex accessibility measures); extent of land use mix-ing; trac calming; street pattern; jobs-housing balance (orland use balance); and pedestrian features such as the avail-ability of sidewalks. Handy et al. (1998) point to the fact thatmany neighborhood-scale studies focus on non-work trips, es-pecially shopping, and how built environment characteristicscan potentially reduce car travelpartly by encouragingwalk-ing as a substitute for driving, and partly by facilitating shortercar trips than would be necessary if a certain facility type werenot available in the neighborhood. e built-environmentfeatures that can encourage walking include objective factors(e.g., how close the facility is to the dwelling, or the avail-ability of sidewalks) as well as more subjective factors (e.g.how pleasant and safe the walking route is perceived to be).Mixed land use is generally considered more conducive to re-ducing car dependency than monofunctional residential zon-ing. is includes local employment opportunities; the con-cept of local jobs-housing balance has been a prevailing plan-ning ideal since the 1970s (Cervero 1989; Roundtable 2008;Weitz 2003).

It is generally assumed that the greater the number of desti-nation choices available within the neighborhood, the greaterthe likelihood that a destinationwithin the neighborhoodwillbe selected. However, asHandy et al. (1998, 10) point out, theavailability of a greater number of choices outside the neigh-borhood, may also increase the likelihood that a destinationoutside the neighborhood will be selected.

Empirical studies suggest that neighborhood characteristicssuch as higher residential densities and the presence of mixedland uses do promote walking as a travel mode in connec-tion with non-work activities (see Boarnet and Crane 2001;Chatman 2005; Handy et al. 2005; Handy and Clion 2001;Handy et al.1998;Rajamani et al.2003). e impacts in termsof reduced vehiclemiles traveled are, however, generally found

to be quitemoderate. Moreover, inmany of the studies of rela-tionships between local built-environment characteristics andtravel behavior, no eort has been made to control for the lo-cation of these neighborhoods within the metropolitan struc-ture, i.e. in relation tomajor concentrations of workplaces andservices. To some extent, a higher local jobs-housing balancehas been found to reduce commuting distances among the res-idents of the areas where new jobs have been established andamong the employees of businesses in the areas where newhousing has been added (Frank and Pivo 1994; Nowlan andStewart 1991). However, this does not necessarily mean thathigher local jobs-housing balances have reduced commutingdistances at a metropolitan scale. Employment growth in pre-dominantly residential suburbsmay result in longer commutesfor those employees who are not local residents. If the work-places in question are specialized and recruit employees froma wide catchment area, this eect may well outweigh any re-duction in commuting distances among the local residents.

3 Case area andmethods

e Copenhagen Metropolitan Area (population: approxi-mately 1.8 million) is one of the largest urban areas in north-ern Europe and a major node for international air and railtransport. Although it now encompasses several smaller citiesthat previously were largely autonomous, the CopenhagenMetropolitan Area is today a conurbation functioning largelyas a single city and as a continuous job and housing market.

According to some authors, historical urban cores have lostmuch of their dominant position during the past 30 to 40years. For example, Sieverts (1999) holds that cities can nolonger be tted into a hierarchic system according to cen-tral place theory and should, instead, be understood as net-works of nodes, where there is a spatially more or less equal,scattered distribution of labor with spatial-functional special-izations. Such net-like cities or city regions have polycen-tric rather than monocentric or hierarchic center structures.However, the Copenhagen Metropolitan Area does not tthis description (which may also be of limited validity in awider European context; cf. Newman and Kenworthy 1989;Nielsen and Hovgesen 2006). e inner city of Copenhagenstill has an unchallenged status as the dominant center of thecity region. e central municipalities of Copenhagen andFrederiksberg, making up only 3.4 percent of the total area ofthe Copenhagen Metropolitan Area, are home to a third ofthe inhabitants and an even higher proportion of the work-places. is implies that people whose residences are locatedclose to downtown Copenhagen travel, on average, consider-

.

ably shorter distances to most facility types than those whoreside in the outer suburbs (Nss 2006b). e center struc-ture of Copenhagen Metropolitan Area could be character-ized as hierarchic, with downtown Copenhagen as the maincenter, the central parts of ve formerly independent periph-eral towns now engulfed by the major conurbation as second-order centers along with certain other concentrations of re-gionally oriented retail stores, and more local center forma-tions in connection with, among others, urban rail stationsand smaller-size municipal centers at a third level.

e study was carried out using a combination of qualita-tive and quantitative research methods. Apart from a regis-tration of urban structural conditions, the data collection in-cluded a large travel survey among inhabitants of 29 residentialareas (1932 respondents), a more detailed travel diary investi-gation among some of the participants of the rst survey (273respondents), and qualitative interviews with 17 households.e questionnaire surveys and the interviews were all carriedout during the period from June to September 2001, avoidingthe main holiday month of July. e chosen residential areas(Figure 1) include a wide range of urban structural situations,varying in their location relative to downtown Copenhagenand lower-order centers, as well as in their density, composi-tion of housing types and availability of local green areas.

e qualitative interviews were (apart from a single case)conducted in the homes of the interviewees, usually lastingbetween 90 minutes and two hours. Nine interviewees werechosen from three inner-city areas (C1, C2, and C4 in Fig-ure 1) and eight were recruited from two outer-suburban ar-eas. Among the latter areas, one is located close to an urbanrail station (La2) whereas the other suburban interviewee area(S4) has very poor accessibility by public transport. All in-terviewswere audio-recorded and subsequently transcribed intheir entirety. e interviews were semi-structured, focusingon the interviewees reasons for choosing activities and theirlocations, travel modes and routes, as well as the meaning at-tached to living in or visiting various parts of the city. As animportant tool for the analysis an interpretation scheme wasdeveloped. By requiring the research team to make writteninterpretations of each interview in light of each of the de-tailed research questions, the interpretation scheme made usread and penetrate the transcribed interview texts more thor-oughly than we would probably have done otherwise.

e questionnaires included questions about respondentstravel behavior, activity participation, socioeconomic charac-teristics, residential preferences, and attitudes to transport andenvironmental issues. e main survey included questionsabout the distances traveled via each mode on each day over

the course of one week. e travel diary investigation cov-ered the four-day period from Saturday morning to Tuesdaynight, and included detailed questions about each trip made(purpose, length, travel time, and travel mode). e concen-tration of respondents in a limited number of selected loca-tions allowed for an in-depth accounting of contextual condi-tions in each of the chosen areas. is enabled us to record alarge number of urban structural characteristics of each resi-dential address within the selected areas and to include thesecharacteristics as variables in the investigation. In total, 38 ur-ban structural variables were measured for each respondent,includingve variablesmeasuring the locationof the residencerelative to theoverall center structure in themetropolitan area,eleven variables indicating the location of the residence in re-lation to rail-bound public transport, seven variables measur-ing the density of the local area and the neighborhood, twelvevariables measuring dierent aspects of service facility accessi-bility in the proximity of the dwelling, two variables measur-ing the availability of local green recreational areas, and onevariable indicating the type of local street structure.

In addition to the urban structural variables, a number ofindividual characteristics of the respondentswere recorded. Inthe subsequentmultivariate analyses, 17 demographic, socioe-conomic, attitudinal, and other non-urban-structural vari-ables were included as control variables:

1. Four variables describing demographic characteristics ofthe respondent and the household to which he or shebelongs (sex, age, number of household members belowseven years of age, number aged 717 years)

2. Seven variables describing socioeconomic characteris-tics of the respondent (workforce participation, stu-dent/pupil, pensioner, personal income, drivers license,two dichotomous variables for type of education)

3. One attitudinal variable (index for transport-related res-idential preferences)

4. Five other control variables indicating particular activ-ities, obligations or social relations likely to inuencetravel behavior during the period of detailed travel reg-istration.

I consider these control variables to be the most relevant among thoserecorded. Analyses have also been carried out with a larger number of con-trol variables. e eects of the main urban-structural variables remainfairly stable across these various analyses. See Nss (2009a) for a discussionof the appropriateness of dierent control variables in studies of relation-ships between land use and travel.

New urbanism or metropolitan-level centralization? Figur 1 Omtrentlig lokalisering af de planlagte undersgelsesomrder, vist med bl cirkler.

P1

T5

T4

T1T3

C5 C1 C6

H1

S1

C2

Li1

Li2

La2La1

C4P2

S2

C3R1

R3

R4

R5

R6

S3

T6

S4

T2

R2

Figure 1: Location of the 29 investigated residential areas. Scale 1:750 000.

A large number of travel behavior indicators were recordedfor each respondent: total traveling distance, traveling dis-tances by car, bus, train, and non-motorized modes, and theproportions of the total distance traveled by car, public trans-port, and non-motorized modes. Each of these eight aspectsof travel behavior was recorded for the weekdays (Monday-Friday), the weekend (Saturday-Sunday) and for the week asa whole, resulting in a total of 24 travel behavior variables.In addition, annual driving distances of any cars belonging tothe household were recorded, as well as ights and other long-distance holiday trips. Due to space constraints, in the follow-ing we shall concentrate on a single travel behavior variable:the distance traveled by car on weekdays. However, the rel-

ative strengths of the inuences of dierent urban structuralvariables on the remaining travel behavior variables are fairlysimilar to the strength of their inuence on traveling distanceby car on weekdays (Nss and Jensen 2005, 353-371).

First, simple bivariate correlations between traveling dis-tance by car and each urban structural variable will be shown.Second, results of analyses where each of these relationshipshas been controlled for the inuences of the 17 non-urban-structural variables as well as the location of the dwelling rela-tive to the city center ofCopenhagenwill be presented. ere-upon, material from our qualitative interviews will be usedin order to explain why metropolitan-level urban structuralvariables exert a stronger inuence on travel behavior than

.

neighborhood-scale urban characteristics do. Finally, the re-sults of an analysis including only the fourmost inuential ur-ban structural variables and the 17 non-urban-structural con-trol variables will be presented.

4 Bivariate correlations

As mentioned in the introduction, neighborhood-scale streetpattern is an urban structural variable oen used in Americanstudies investigating relationships between urban built envi-ronment and travel behavior. Compared to the curvilinearand cul-de-sac street patterns typical for suburban neighbor-hoods planned according tomodernist principles, grid-shapedstreet networks facilitate more direct access to local destina-tions and can thus bring a larger number of local facilitieswithin acceptable walking (or biking) distance (Handy et al.1998). Street patterns in the neighborhood were recordedin the Copenhagen Metropolitan Area too. Among the 29investigated residential areas, nine were located in neighbor-hoods characterized by a (more or less) grid-shaped street pat-tern, whereas the remaining 20 residential areas were locatedin neighborhoods characterized by other kinds of street pat-terns (curvilinear streets or some sort of hierarchical street lay-out based on recommended driving speeds, with cul-de-sacaccess roads and separation of motorized and non-motorizedtrac). Figure 2 shows an example of investigated neighbor-hoods with street patterns belonging to the grid and the non-grid categories.

In line with what has been found in several American stud-ies (e.g. Cervero 2003) (e.g. Cervero, 2003; Frank, 2003), theamount of car travel is lower among residents living in neigh-borhoods characterized by grid-shaped than by other types ofstreet patterns. As can be seen in Figure 3, mean traveling dis-tance by car during the weekdays from Monday through Fri-day is only 86 km among the respondents living in a neigh-borhoodwith grid-shaped street pattern, compared to 153 kmamong the respondents living in local areas with other types ofstreet

However, the correlation between street pattern andamount of car travel does notnecessarily reect any causal rela-tionship. Admittedly, grid-like street patternsmay oer betterconnectivity between dierent locations within the neighbor-hood and may thus facilitate shorter local traveling distances,especially compared to cul-de-sac-based street patterns. eshorter local traveling distancesmay also be conducive to non-motorized travel, since people who rely on their own musclesfor transportation are usually sensitive to travel distance andoen change to motorized modes if the distance exceeds a

Figure 2: Examples of investigated residential areas with grid andnon-grid street patterns. Amager North (top) andHolmene (bottom).

comfortable level. Nevertheless, the dierences in travelingdistances resulting from the local street patterns are probablynot very large, since the investigated neighborhoods are them-selves of a limited size. Arguably, the location of the neigh-borhoods relative to concentrations of workplaces and otherfacilities matters more. Figures 4 and 5 provide some prelimi-nary indication of this. Here, the respondents have been sub-divided into four groups of approximately equal size, depend-ing on how far from the Copenhagen city center they live. Aswe can see from Figure 4, the respondents living far from thecity center travel, on average, considerably farther by car thantheir counterparts living in the inner distance belts, especiallythose living less than six kilometers away from the city cen-ter. Among the latter group of respondents, the mean trav-eling distance by car over the ve weekdays is 66 km, com-pared to 176 km among respondents living more than 28 kmaway from the city center of Copenhagen. Figure 5 showsmean trip distances for journeys to work/education as well


Street pattern in the neighborhood of theresidence

Non-grid street patternGrid street pattern

Mean

trave

ldist

ance

byca

rMon

day-

Frida

y(km

)

200

150

100

50

0

Figure 3:Mean traveling distances by car during the weekdays(Monday-Friday) among respondents living in residentialareas with grid and non-grid street patterns. N = 1810, ofwhich 603 in grid-type and 1207 in other street environ-ments.

as several other trip purposes. e same pattern as for over-all traveling distances by car are evident for the various cate-gories of trips: suburbanites tend to make longer trips thaninner-city dwellers do. On average, our respondents (includ-ing both workforce participants and non-participants) made3.4 journeys to work or education during the week, 3.5 shop-ping/errand trips, 1.0 trip for bringing/picking up children,1.3 visiting trips and 2.4 leisure trips. Since journeys to workare, on average, longer than trips for any of the other purposes,these gures imply thatmuch of the dierence between inner-city dwellers and suburbanites in overall traveling distances(and traveling distances by car) is attributable to the longercommuting distances among respondents living in the outerparts of the metropolitan area.

Table 1 shows the bivariate relationships of each of the 38urban structural variables with the respondents distance trav-eled by car on weekdays, as well as partial correlations of eachurban structural variable with the distance traveled by car,

Trips longer than 100 km have been excluded. Trip distances for jour-neys to work are based on workplace locations given by workforce partic-ipants who responded to the main survey, with distances measured alongthe road network by means of the GIS program ArcView. Trip distancesfor other purposes are based on the travel diary survey (273 respondents).e gures displayed in the diagram for non-work trips are weighted aver-ages of trip lengths during weekdays and weekends.

Trips home from any of these destinations are not included in thesegures.

Distance from the dwelling to the city center ofCopenhagen (km)

over 2815 - 286 - 15below 6

Mean

trave

lingd

istan

ceby

carM

onda

y-Frida

y(km

)

200

150

100

50

0

Figure 4:Mean weekday (Monday-Friday) traveling distances by caramong respondents living within dierent distance inter-vals from the city center of Copenhagen. N = 1810, vary-ing from 405 to 530 per distance belt.

Below 6 6 to 15 15 to 28 Over 280

5

10

15

20

25

sub-title

Journey to work (N = 1234) Shopping/errand (N = 585)Bringing/picking up children (N = 123) Visiting trips (N = 264)Leisure trips (N = 421)

Distance from dwelling to the city center of Copenhagen (km)

Mea

n tr

ip le

ngth

(km

)

Figure 5:Mean one-way trip lengths for dierent purposes amongrespondents living within dierent distance intervals fromthe city center of Copenhagen.

controlling for the location of the dwelling relative to the citycenter ofCopenhagen and17non-urban-structural variables.

ese control variables are: sex, age, number of children younger thanseven years of age in the household, number of children aged 717 in thehousehold, personal income, possession of a drivers license for car, whetheror not the respondent is aworkforce participant, whether or not the respon-dent is a student, whether or not the respondent is a pensioner, whether or

.

Table 1: Relationships between various metropolitan-scale and neighborhood-scale urban structural variables and the respondents travelingdistance by car on weekdays (Monday- Friday).

Bivariatecorrelations(Pearsons r)

Partialcorrelations

Location of the dwelling relative to the metropolitan-level center structure:Location of the dwelling relative to downtown Copenhagen (non-linear function) 0.233***Linear distance along the road network from the dwelling to downtown Copenhagen 0.201***Logarithmic dist. along the road network from the dwelling to the closest second-order center 0.210*** 0.092***Linear dist along the road network from the dwelling to the closest second-order center 0.177*** 0.056*Logarithmic dist. along the road network from the dwelling to the closest regional shopping mall 0.166*** 0.019Location of the dwelling relative to rail stations:Logarithmic dist. along the road network from the dwelling to the closest urban rail station 0.199*** 0.072**Linear dist. along the road network from the dwelling to the closest urban rail station 0.164*** 0.062*Linear dist. along the road network from the dwelling to the closest well-serviced junction station 0.159*** 0.027Linear dist. along the road network from the dwelling to the closest junction station 0.151*** 0.053*Linear dist. along the road network from the dwelling to any rail station 0.169*** 0.097***Residential location less than 500 m away from closest urban rail station (yes = 1, no = 0) 0.094*** 0.019Residential location less than 1000 m away from closest urban rail station (yes = 1, no = 0) 0.124*** 0.016Residential location less than 500 m away from any rail station (yes = 1, no = 0) 0.089*** 0.022Residential location less than 1000 m away from any rail station (yes = 1, no = 0) 0.102*** 0.007Residential location less than 500 m away from closest junction station (yes = 1, no = 0) 0.113*** 0.034Residential location less than 1000 m away from closest junction station (yes = 1, no = 0) 0.117*** 0.036Density in the surroundings of the dwelling:Density of inhabitants & jobs in the local area of the dwelling (inhab.+jobs within a radius of 800 m) 0.203*** 0.071**Population density in the local area of the dwelling 0.191*** 0.048Job density in the local area of the dwelling 0.188*** 0.075**Density of inhabitants and jobs in the narrowly demarcated residential area 0.231*** 0.053*Population density in the narrowly demarcated residential area 0.234*** 0.054*Dwellings per hectare in the narrowly demarcated residential area 0.230*** 0.050*Job density in the narrowly demarcated residential area 0.120*** 0.019Availability of service facilities in the proximity of the dwelling:Combined index for availability of service facilities in the proximity of the dwelling 0.234*** 0.107***Index for availability of shopping opportunities in the proximity of the dwelling 0.217*** 0.095***index for availability of primary schools, kindergartens and crches in the proximity of the dwelling 0.192*** 0.084***index for availability of public-sector oces in the proximity of the dwelling 0.175*** 0.075**Number of grocery stores within 1.5 km distance from the dwelling 0.204*** 0.077**Number of special commodity stores within 1.5 km distance from the dwelling 0.171*** 0.065*Linear dist. along the road network from the dwelling to the closest grocery store 0.164*** 0.075**Linear dist. along the road network from the dwelling to the closest post oce 0.122*** 0.044Linear dist. along the road network from the dwelling to the closest town hall 0.160*** 0.070**Linear dist. along the road network from the dwelling to the closest primary school 0.156*** 0.065*Linear dist. along the road network from the dwelling to the closest kindergarten 0.139*** 0.01Linear dist. along the road network from the dwelling to the closest crche 0.233*** 0.144***Local green recreational areas:Availability of a recreational area of at least 10 hectares within 0.5 km distance from the dwelling 0.041 0.043Availability of a recreational area of at least 10 hectares within 1 km distance from the dwelling 0.041 0.042Local street pattern:Grid structure (1) or other street patterns (0) 0.189*** 0.004


All the bivariate correlation coecients have the expectedsigns: Travelingdistances by car arehigher among respondentsliving: peripherally in relation to various types of centers in thecenter structure of Copenhagen Metropolitan Area; in areaswith a modest supply of local facilities; far away from urbanrail stations; and in areas where population density and work-place density are low. ere is also a (rather weak) tendencytoward increased car travel among respondents who have pooraccess to local green recreational areas.

In line with the impression le by Figure 4, we nd a quitestrong bivariate correlation between distance traveled by carand location of the dwelling relative to the city center ofCopenhagen, especially when the distance to the city center istransformed by means of a non-linear function (r = 0.233).Some local-scale urban structural variables also show strongcorrelations with the amount of car travel, notably densitieswithin the narrowly demarcated residential area, measured ei-ther as population density (r =0.234), density of dwellings(r = 0.230) or as a combined population and job densityvariable (r = 0.231). ere are also correlations of similarstrength between the amount of car travel and, respectively,the distance from the dwelling and the distance to the closestcrche provider (r = 0.233) and a combined index for avail-

not the respondent has completed advanced studies (more than six years fol-lowing the standard nine years of primary school) in a technical eld suchas engineering or economics, whether or not the respondent has a short ormedium-long education (less than six years following primary school) asa tradesman or industrial worker, transport-related residential preferences,whether or not the respondent hasmoved to the present dwellingwithin theprevious ve years, employment-related trips during the investigated week,overnight stays away from home more than three nights, number of days atthe workplace or school during the investigatedweek, and regular transportof children to school or kindergarten.

e city center was dened as the City Hall Square. Based on the-oretical considerations as well as preliminary analyses of the empiricaldata, the location of the residence relative to the city center of Copen-hagen was measured by means of a variable constructed by transform-ing the linear distance along the road network by means of a non-linearfunction. is function was composed of a hyperbolic tangential func-tion and a quadratic function, calculated from the following equation:AFSTFUN = ((EXP(centafs0.18 2.85)) EXP( (centafs0.18 2.85))) / (EXP(centafs0.18 2.85) + EXP( (centafs0.18 2.85))) (0.00068( centafs 42)(centafs 42) 2.8) where ASTFUN = thetransformed distance from the dwelling to the city center and centafs = thelinear distance along the road network, measured in km.

Crche describes childcare for children under the age of three years.

ability of service facilities in the proximity of the dwelling(r =0.234).

As can be seen in Table 1, all urban structural variablesshow statistically signicant bivariate relationships with theamount of car travel (p < 0.001) except the two variablesindicating availability of green outdoor areas in the proxim-ity of the dwelling. Local street pattern (see Figure 4) showsa fairly strong bivariate correlation with distance traveled bycar (r = 0.189), but not as strong as with the location ofthe dwelling relative to the city center of Copenhagen. Cartravel is also quite strongly correlated with logarithmic dis-tance from the dwelling to the closest second-order center,local-area density of inhabitants and jobs, and logarithmic dis-tance from the dwelling to the closest urban rail station.

5 Controlling for non-urban-structural variablesand the location of the dwelling relative to thecity center of Copenhagen

Controlling for the location of the dwelling relative to the citycenter of Copenhagen as well as for the non-urban-structuralvariables (Table 1, right column) considerably weakens thecorrelations of many of the neighborhood-scale urban struc-tural variableswith car travel. is is particularly true for streetstructure; the correlation between it and the amount of cartravel becomes insignicant when we control for residential

e service index was constructed as a weighted sum of z-scores forthree other indices also included in the analysis: an index for availability ofshopping opportunities near the residence, an index for availability of pri-mary schools, kindergartens, and crches in the environs of the dwelling,and an index for availability of public-sector oces in the proximity of thedwelling. e weighting between the sub-indices was based on data fromNorwegian (Vibe 1993, 35) and Danish (Christensen 1996, 9) nationaltravel surveys on the frequencies of dierent trip purposes. Given a weightsum of 100, this implied the following weights: Shopping opportunities,60; schools, kindergartens and crches, 27; and public oces, 13. e threelatter indices were constructed as follows: e shopping opportunity in-dex was constructed by adding the z-scores for number of grocery storeswithin 1.5 km of the dwelling, number of special commodity stores within1.5 kmof the dwelling, anddistance from the dwelling to the closest grocerystore (with the sign of the z-score changed for the latter factor in order tomake a high index value signify a high accessibility for all the sub-elementsof the index). e index for availability of primary schools, kindergartensand crches was based on the measured distances from the dwellings to theclosest facilities in these categories, with a weighting based on considera-tions about the number of years the children spent in each type of institu-tion and the propensity of parents to follow their children to the varioustypes of institutions. Again, the sign of the index value was changed in or-der tomake a high index value signify high accessibility. e index for avail-ability of public-sector oces was constructed in a similar way, based on theassumption that residents go to the town hall to make use of public servicesabout as oen as they visit the post oce.

.

location relative to the city center. In this study, all the inves-tigated residential areas having a grid-like street pattern are lo-cated in the inner part of theCopenhagenMetropolitanArea,no more than 9 km from downtown. e correlations of cartravel with the location of the dwelling relative to lower-ordercenters as well as with the various local-area and residential-area density variables are alsoweakened, but these correlationsare still statistically signicant, at least when density is mea-sured within a local area larger than the narrowly demarcatedresidential area.

e weakened correlation coecients of severalneighborhood-scale variables reect the fact that thereare considerable internal correlations between the dierenturban structural variables. Population densities, workplacedensities, accessibility to public transport, and availabilityof service facilities near the residence are generally higherthe closer to the city center of Copenhagen the residenceis located (see Table 2). Many of the bivariate correlationsbetween neighborhood-scale urban structural variables andcar travel are therefore attributable to the location of theseneighborhoodswithin the overall urban structure, rather thanto the proximity of each residence to local service facilities.

e various service index variables also have statistically sig-nicant correlations with the amount of car travel when non-urban-structural variables and the location of the dwelling rel-ative to the city center are taken into account. e correla-tion between car travel and the distance from the dwelling tothe closest crche is especially strong. However, proximity toa crche is hardly an important determinant of the amount ofcar travel among the respondents. Only aminority among therespondents follow children to and fromcrches, and the aver-age length of such trips is considerably shorter than, for exam-ple, journeys to work. Rather, this variable serves as a proxyfor other urban-structural conditions, notably proximity to asecond-order or third-order center wheremany other facilitiesthan crches are available. Moreover, as there is a strong cor-relation between the distance from the dwelling to the clos-est crche and the distance to the city center, the former vari-able may steal some of the eect of the latter variable whenboth are included in the same analysis. ese two variablesare equally strongly correlated with the amount of car travelwhen control is made for non-urban-structural variables, andthe theory-blind way that the statistical soware used in this

Among the respondents of our main survey, 62 percent reported tripsto a workplace or place of education at least four times during the week ofinvestigation, whereas only 12 percent reported regularly bringing childrento and from kindergarten, childcare, or school. In addition, mean commut-ing distances are more than four times as long as trips from home to kinder-garten/crche.

research carries out the multivariate calculations implies thatthere is always a risk that a theoretically less-well- founded vari-able may mask some of the relationship between a dependentvariable and a theoretically more plausible independent vari-able.

Needless to say, the location of the dwelling relative to thecity center of Copenhagen has not been included in the ta-ble, as this is one of the control variables. However, if we in-stead use the variable showing the second-strongest relation-shipwith car travel when controlling for non-urban-structuralvariables (i.e. the combined index for availability of service fa-cilities in the proximity of the dwelling) as the urban struc-tural control variable, the correlation coecient between theamount of car travel and the location of the dwelling relativeto the city center of Copenhagen is still as high as 0.132, witha level of signicance of 0.000.

With the exception of proximity to crches, the local-scaleurban-structural variables tend to become less strongly corre-lated with car travel when the analysis includes control for thelocation of the dwelling relative to the city center of Copen-hagen. In the next section, we shall draw on qualitative inter-views data that may help explain why this is so. We also note(in the right column of Table 1) that some of the service in-dex variables are more strongly correlated to car travel is morestrongly correlated to some of the service index variables thanto local area density or to the location of the dwelling relativeto second-order centers. In the analyses presented in Section5, we have still preferred to include the latter variables insteadof any of the service indexes. is is because we believe thatthe service indexes act, to some extent, as proxy variables forthe latter variableswhich are, in our view, more appropriateas the centers include not only concentrations of service fa-cilities but also concentrations of workplaces, and oer highaccessibility by public transport.

6 Rationales inuencing travel behavior

Why does travel behavior in the Copenhagen MetropolitanArea depend more on metropolitan-scale than on local-scalebuilt environment characteristics? Material from the qual-itative interviews illuminates some important rationales onwhich people base their travel behavior. e relative impor-tance of metropolitan-scale and neighborhood-scale built en-vironment characteristics to travel behavior depends, in par-ticular, on peoples rationales for choosing the locations of theactivities in which they participate.

e interviewees choices of locations for their activities aremade as a compromise between two competing desires: the


Table 2:Mean values for some urban structural characteristics of the respondents residences, grouped into four distance intervals from the citycenter of Copenhagen. N = 1932 respondents of the main survey.

Urban structural factor Distance interval from the city center of Copenhagen28 km(N=457)

Distance from residence to downtown Copenhagen (km) 3 9.7 21.9 41.8Distance from residence to closest second-order urban center (km) 2.2 5.7 7.6 10.1Distance from residence to closest urban rail station (km) 1.4 2.3 3 11.2Local area population density (inhabitants/ha) 85 24 14 10Local area workplace density (jobs/ha) 66 11 7 5Distance from residence to closest grocery store (km) 0.13 0.51 0.97 0.68Number of grocery stores within 1.5 km distance from the dwelling 150 18 12 8Number of specialized stores within 1.5 km distance from the dwelling 218 16 15 15Distance from residence to closest primary school (km) 0.51 0.85 1.5 1.2Distance from residence to closest kindergarten (3-5 years) (km) 0.31 0.52 0.65 0.77Distance from residence to closest crche (0-3 years) (km) 0.32 0.72 1.08 2Distance from residence to closest post oce (km) 0.74 1.09 1.56 0.98Distance from residence to closest town hall (km) 2.8 2.8 4 3.7Proportion of residences with a green recreational area of at least 10 ha within 1km distance (%)

36 60 55 42

desire to limit travel distances and the desire for the best fa-cility. Depending on the trip purpose, the balance betweenthese desires may vary somewhat. Our interviews suggest thateach resident establishes an individual threshold value for thelongest acceptable travel distance within each category of des-tination.

Among the dierent travel purposes, travel related to em-ployment and higher education, along with visits to friendsand relatives, are the purposes for which the longest distancesare accepted. Because the workplace or school/university isusually visited each weekday, while long trips for social pur-poses are far less frequent, journeys to work or education ac-count for the largest proportionof the travel distance onweek-days. e acceptable travel distance to work or education ap-pears to be greater for travelers with more specialized workqualications, those withmoremobility resources at their dis-posal, and those living further from the largest concentrationsof work and education opportunities.

For example, one interviewee, a computer engineer now liv-ing in the peripheral suburb ofUvelse, told that his present jobwas chosen without much consideration of the distance fromtheir dwelling at that time:

No, I just thought that I wanted employment,and then we would have to see where to go. And itdid not matter where the workplace was located.(Male computer engineer living in the suburb ofUvelse, 30 years old.)

Similarly, an economist living in the same suburb reportedthat he preferred to commute all the way to downtownCopenhagen instead of nding a less-challenging job in a mu-nicipality closer to the residence:

Surely, I would like something closer to home,but there are no such [relevant] jobs available tome here in the vicinity. en it would have to beif I were interested in working in a municipal ad-ministration. ... But that would be such a smallworkplace, and I simply want some more chal-lenges. ... Yes, for sure, mostwork opportunities foreconomists are in the city of Copenhagen. (Maleeconomist, living in the suburb of Uvelse, 38 yearsold.)

One of the interviewees living in the inner-city area of Fred-eriksberg reported that both his own and his wifes work-places were chosen primarily because they found the jobs in-teresting. Both had quite specialized job skills (civil engineerand pharmacology researcher). Due to the central location oftheir dwelling, they still succeeded in nding satisfactory jobswithin a moderate distance from home. Asked whether theywould have taken these jobs if theworkplaceswere located fur-ther away, for example in Roskilde (40 km away), the intervie-wee replied:

You see, we have never made long journeys towork. ... It would of course depend on the situa-

.

tion. If it was not possible to get a job closer tohome, yes. ... Roskilde is pretty easily accessible, butyou see ... I dont think that I would be willing todrivemore than half an hour or so to get there, I re-ally dont. (Male civil engineer, living in the inner-city area of Frederiksberg, 43 years old.)

Since primary schools, kindergartens and well-stocked gro-cery stores can usually be found closer to the residence thancan jobs matching specialized work qualications, the thresh-old values for acceptable distances to such facilities are usuallyshorter than for workplaces and higher education opportuni-ties.

Distance limitation is included as an important (butnot theonly) rationale for most interviewees choices of locations fordaily-life activities. e desire to limit travel distances may begrounded on dierent reasons, oen in combination, such assaving time, saving money, personal physical limitations withrespect to walking and bicycling, and the desire to support thelocal community and maintain local social contacts.

Along with distance limitation, a desire for the best facil-ity (judged against the instrumental purpose of the trip) is themost important rationale for the interviewees choices amongdestinations. In a way, this is the most fundamental rationale,as the trips would simply not occur if no suciently attractivefacility existed that might be visited. In practical locationalchoices, distance limitations and the desire for the best facil-ity must be weighed against each other. What is consideredthe best facility will vary with the purpose of the trip andwith the individual characteristics of the person in question.For workplaces, factors like job duties, qualication require-ments, wages, and work environment are relevant. For spe-cialized jobs, the catchment area from which employees arerecruited typically includes considerable parts of the region.

Factors inuencing the perceived quality of shops include,among others, the selection of goods available, prices, and pos-sibly the availability of parking. When living in the periphery,the local grocer is oen used only for emergency purchases,e.g. when there is no coee le in the house. Among those liv-ing in the central parts of Copenhagen, local shops are oenwell- stocked and are relied on to a higher extent for ordinaryshopping. Among kindergartens, the reputation of the insti-tution (pedagogy, etc.) and perhaps also the ethnic composi-tion of the children may be inuential factors.

e tendency to choose a facility other than the closest oneis indicated in Figure 6, where lengths of tripsmade by parentswhen transporting their children to kindergarten or crcheare compared to the distances from dwellings to the closestsuch facilities. Although parents usually do not choose (nor

are they oered places for their children in) kindergartens andcrches located very far away from the family dwelling, the fa-cilities actually used are, on average, located more than fourtimes farther from the home than the nearest kindergartensand crches. We still see a clear tendency toward longer tripsto such facilities among respondents living farther from thecenter of the metropolitan area, reecting the higher densityof facilities in the inner city than in the outer suburbs.

Below 6 6 to 15 15 to 28 Over 280

1

2

3

4

5

6

7

8

Mean distance to closest kindergarten and creche (N = 1932)Mean trip distance for bringing children on weekday mornings (N = 33)

Distance from dwelling to the city center of Copenhagen (km)

km

Figure 6:Mean trip distances for trips with the purpose of trans-porting children on weekday mornings (Monday-Tuesday)among travel diary survey respondents living within dif-ferent distance intervals from the Copenhagen city center,compared to distances from home to the closest kinder-gartens and crches among main survey respondents livingin the same distance intervals. N = 33 travel diary tripsand 1932 main survey respondents.

e destinations of visiting trips are dened entirely by thetravelers family relations and circle of acquaintances. Whenit comes to leisure trips, the choice among facility categoriesdepends strongly on the interests and lifestyle of the person inquestion, but quality dierences within each facility categorymatter as well. For example, when planning an outing, a forestthat is distant but larger and more beautiful may be preferredto a local forest.

e trip distances shown in Figure 6 are based on the follow-up traveldiary survey, in which 273 of the original 1932 respondents participated.Among a total of 231 visiting trips carried out during the period Satur-dayTuesday, only trips on Monday and Tuesday were included, and onlytrips carried out as the rst trip of the day, thus including only trips origi-nating at the residence. When measuring average distances from dwellingsto the closest kindergarten and crche, the distances have been weighted inorder to take into account the fact that trips to bring children to kinder-gartens are carried out about twice as frequently as trips to bring them tocrche.


Besides emphasizing the possibility of choosing the instru-mentally best facility, the atmosphere and aesthetic quali-ties of the destination are important to many of our intervie-wees. In particular, this applies to trips such as visits to restau-rants, cinemas, theaters, and other cultural facilities, as well asto shopping (particularly when purchasing non-grocery com-modities). In contrast, peoples choices of locations in whichto seek employment are inuenced to a much lesser extent bythe atmosphere of the district where the workplace is lo-cated.

In quotidian travel, some trips are more fundamental, andtheir characteristics more xed, than other trips. Oen, suchtrips are part of a trip chain. Other travel purposes are thenlinked to this fundamental trip. For example, by choosinga well-stocked store along the route followed anyway on theway home from work, the rationale of distance limitation canbe combined with the rationale of choosing the best facility.is can, to some extent, compensate for the longer distancesto shops typical of residences in the outskirts of the city. iskind of adaptation is very common among our interviewees.

Formost travel purposes, our respondents and intervieweesemphasize the possibility of a choice of facilities over proxim-ity. is means that the amount of travel is inuenced to agreater extent by the location of the residence in relation toconcentrations of facilities than by the distance to the closestsingle facility within a category. is is particularly evidentfor workplaces and places of higher education, but also forcultural and entertainment facilities, specialized stores and, tosome extent, grocery stores. For leisure activities, the atmo-sphere and the aesthetic qualities of a destination may alsoplay a role.

us, formost travel purposes, our interviewees donot nec-essarily choose the closest facility, but rather travel a bit fartherif they can then nd a better facility. ey tend to emphasizea rationale of choosing the best facilities above a rationale ofminimizing the friction of distance. is is especially true inregard to workplaces. Travel distances therefore depend moreon the location of the dwelling relative to large concentrationsof facilities than on the distance to the closest facilities. Peo-ple who live close to the city center can access a large num-ber of facilities within a short distance from the dwelling andtherefore do not have to travel long distances, even if they arevery selective as to the quality of the facility. Since the largestconcentrations of workplaces, as well as other facilities, are inthe city center and the inner districts of the city, the above-mentioned circumstances imply that the amount of quotidiantravel is inuenced by how far away the interviewees live from

the city center rather than by the distance from their dwellingto lower-order centers.

Table 3 summarizes the impacts of the rationales identiedin the qualitative interviews on relationships between residen-tial location and travel . ese eects are attributable to theability of the rationales to inuence the relationships betweenresidential location and concentrations of facilities at both lo-cal and metropolitan scales. e rationale of distance limi-tation has been divided into two aspectslimitation of geo-graphical distance and limitation of time usagebecause theyappear to inuence travel behavior dierently.

e relationship between the amount of travel and the dis-tance from the residence to the main center of the urban re-gion is strengthened, in particular, by the rationale of beingable to choose the best facility (judged against the instrumen-tal purpose of the trip). e rationales of limiting geographicaldistance and limiting travel time also contribute to this rela-tionship, to some degree, because the regions largest concen-tration of facilities serves as a local concentration of facilitiesfor a large number of inner-city residents in the regions majorcity and because the urban center approximates the geograph-ical center of gravity even for themore peripheral destinationsthat mightfrom a rationale of time-savingbe chosen bycar drivers who want to avoid congested streets. e rationaleof enjoying atmosphere and aesthetic qualities also increasesthe inuence of the distance between the residence and down-town on the amount of travel. e relationship between theamount of travel and the distance from the residence to localfacilities is based, rst and foremost, on the rationale of limit-ing geographical distances; however, it is also aected by therationale of saving time, as the local facilities will oen be theones that can be reached most quickly.

Onemight perhaps imagine that the rationales of inner-cityresidentswould dier from those of suburban residents. How-ever, no clear dierences of this type emerged. e rationaleswere fairly similar across residential locations, but the need toemphasize one rationale at the cost of another was much lesspresent among inner-city residents than among suburbanites.

Because their trips are oen short, inner-city residents alsomake a higher proportion of trips by bicycle or on foot. is,and the generally higher level of public transport service inthe inner city, helps to reduce the amount of car travel amonginner-city dwellers.

e rationales appear to be of a high generality across cultural and so-cial contexts. For example, very similar rationales for location of activi-ties were found in a study of residential location and travel in HangzhouMetropolitan Area, China (Nss 2009a).

.

Table 3:e contributions of the various rationales for location of activities to the relationships between residential location and travel.

Rationales for activitylocation

Frequency of occurrence Inuence on the relationshipbetween the amount of traveland the distance from thedwelling to the main centerof the metropolitan area

Inuence on the relationshipbetween the amount oftravel and the distance fromthe dwelling to local facilities

Limitation of geographicaldistances

Emphasized by allinterviewees, in particularthose without a car.resholds for acceptabledistances vary betweenactivity types and betweenindividuals

Contributes to some extentto this relationship, bothbecause the facilities indowntown Copenhagen arethe closest opportunities forinner-city residents, andbecause of the shortage offacilities in the periphery

Contributes strongly to thisrelationship by increasingthe likelihood of choosinglocal facilities rather thanmore distant ones

Limitation of timeconsumption

Emphasized by allinterviewees, but thresholdsfor acceptable timeconsumption vary betweenactivity types and betweenindividuals

May induce some car driversto choose, e.g., suburbanshopping malls instead ofcentral-city shops.Contributes nevertheless tosome extent to therelationship between thedistance from the residenceto downtown and theamount of travel, due to thefunction of the urban centeras geographical point ofgravity

Contributes to thisrelationship because it willusually take a short time togo to local facilities. Butbecause travel speeds willoen be higher when goingto e.g. a more distantshopping mall with ampleparking space, the inuenceof this rationale is not asstrong as the inuence of therationale of limitinggeographical distances

Wish for the best facility(judged against theinstrumental purpose of thetrip)

Emphasized by allinterviewees, but itsimportance varies betweenactivity types and betweenindividuals

Contributes strongly to thisrelationship by increasingthe likelihood of traveling tothe large concentration offacilities in the inner parts ofthe metropolitan area, butalso because of downtownsrole as a point of gravity forall peripheral destinations.

Contributes to a certainweakening of thisrelationship by increasingthe likelihood of choosingdistant facilities rather thanlocal ones

Enjoying atmosphere andesthetic qualities

Emphasized by manyinterviewees, primarily fornon-bounded trips

Contributes to thisrelationship by directing ahigher number of non-worktrips to the historical urbancore

May contribute to a certainweakening of thisrelationship by makingrespondents bypass facilitiesin local centers where theatmospheric qualities arelower than in the downtownarea


7 Multivariate analyses with mainurban-structural variables andnon-urban-structural control variables

Based on theoretical considerations, the information from thequalitative interviews about the interviewees rationales for se-lecting activity locations, and preliminary analyses of the cor-relations of individual urban-structural variables with travelbehavior, the following four urban-structural variables wereincluded in the main statistical analyses of this study:

Location of the residence relative to the Copenhagencity center

Location of the residence relative to the closest second-order center

Location of the residence relative to the closest urban railstation

Density of inhabitants and workplaces in the local areasurrounding the residence

ese urban-structural variables oer the greatest explanatorypower for interpreting travel behavior variables. Limiting thenumber of urban-structural variables to these four avoidedproblems of multicollinarity.

Table 4 shows the results of amultiple regression analysis offactors potentially inuencing the distance traveled by car onweekdays, including the four above-mentioned urban struc-tural variables and the seventeen non-urban-structural con-trol variables used in the analyses presented earlier. Eectsmeeting the required signicance level are evident for all foururban-structural variables. e eect of the location of the res-idence relative to theCopenhagen city center is, however, con-siderably stronger andmore certain (= 0.130, p = 0.0001)than the other three urban structural variables (absolute val-ues ranging from 0.048 to 0.062, with p values ranging from0.07 to 0.08).

e strong eect of the location of the dwelling relative tothe city center does not, of course, imply that the city center it-self (i.e. CityHall Square) is the destination of a large number

If only variables meeting a signicance level of 0.05 are allowed to beincluded in the model, the location of the dwelling relative to the closestsecond-order center is excluded. e standardized regression coecientsand p-values of the remaining three urban structural variables are thenas follows: Location of the residence relative to downtown Copenhagen: = 0.133, p = 0.0000; local area density: = 0.088, p = 0.0048;logarithmic distance from the residence to the closest urban rail station:= 0.067, p = 0.0067.

of trips. e trip destinations reected in the eect of residen-tial location relative to the city center are the numerous work-places and other facilities concentrated in and around the citycenter. In this sense, distance to the center is a proxy for othercharacteristics. e important point is that the areas centrallocation supports the high concentration of facilities in thisparticular part of the metropolitan area. Centrality implies ahigh concentration of facilities, and vice versa.

Based on the results shown in Table 4, Figure 7 shows (bymeans of black dots) how the expected traveling distancesamong respondents living within each of the 29 residential ar-eas varies with the distance from the residential area to the citycenter of Copenhagen when controlling for the non-urban-structural variables in the regression model. Expected traveldistance by car over the ve weekdays is nearly four times aslong (187 km) in the most peripheral area investigated thanin the most central area (50 km).

In addition, the triangles in Figure 7 illustrate the relation-ships between traveling distances by car and the distance fromthe residential area to the city center of Copenhagen whencar ownership and attitudes toward car traveling are added tothe other control variables. Several studies have shown thatcontrolling for car ownership and attitudes toward car travelreduces the eects of urban-structural variables. As can beseen, this also applies to the Copenhagen case. ere is stilla fairly strong and statistically certain eect of residential lo-cation on the amount of car travel (p = 0.000), with pre-dicted traveling distances twice as long in the outer suburbs asin the inner city. Moreover, our ndings show that car own-ership and, to some extent, transport attitudes are both inu-enced by residential location (see Nss 2009b for a thoroughaccount). Treating car ownership and attitudes to car travelas exogenous control variables not inuenced by urban struc-ture will, therefore, lead to an underestimation of the impactsof residential location on travel. As long as socio-demographicvariables and transport-related residential preferences have al-ready been controlled for, it is my opinion that car ownershipand attitudes to car travel should not be included as additionalcontrol variables. I therefore consider the black dots in Fig-ure 6 to provide a more appropriate representation than thetriangles of the inuence of residential location on travelingdistances by car.

As can be seen in the diagram, expected amounts of cartravel in some of the residential areas are higher or lower thanwhat would be the case if location relative to Copenhagenscity center was the only urban-structural variable inuencingthe amount of car travel. For example, two residential ar-eas located about 35 km from the city center have consider-

.

Table 4: Results from amultivariate analysis of the inuence of various independent variables on the distance traveled by car (km) onweekdays.Only variableswith a level of signicance of 0.15or lower are included. N = 1564 respondents from29 residential areas inCopenhagenMetropolitan Area. Adjusted R2 = 0.272.

Unstandardizedcoecients

Standardizedcoecient

Level ofsignicance(p-value,

two-tailed test) Std. error

Occupational trips during the investigated week (yes = 1, no = 0) 85.65 9.54 0.21 0.0000Index for residential location preference (1 = preference for residential loca-tion facilitating public transport, walking or biking, 0 = no such preferenceexpressed)

47.14 7.22 0.143 0.0000

Possession of a drivers license for car (yes = 1, no = 0) 63.22 11.06 0.131 0.0000Location of the residence relative to downtownCopenhagen (non-linear dis-tance function, values ranging from 0.66 to 3.80)

17.28 4.32 0.13 0.0001

Personal annual income (1000 DKK) 0.088 0.019 0.119 0.0000Overnight stays away from home more than three nights (yes = 1, no = 0) 53.56 13.41 0.087 0.0001Long technical or economic education (yes = 1, no = 0) 44.96 12.02 0.086 0.0002Workforce participation (yes = 1, no = 0) 26.73 9.18 0.07 0.0037Short or medium-long education as a tradesman or industrial worker (yes =1, no = 0)

29.47 10.61 0.064 0.0056

Density of inhabitants and workplaces within the local area of the residence(inhabitants. + workplaces per hectare)

0.168 0.093 0.062 0.0699Logarithm of the distance (meters) from the residence to the closest second-order urban center (log values ranging from 2.49 to 4.46)

24.73 14.11 0.055 0.0799

Sex (female = 1, male = 0) 17.88 7.82 0.054 0.0224Logarithm of the distance (meters) from the residence to the closest urbanrail station (log values ranging from 1.90 to 4.47)

14.18 8.03 0.048 0.0776

Number of household members below 7 years of age (p > 0.15)Age (deviation from being middle-aged, logarithmically measured) (p > 0.15)Number of days at the workplace or school during the investigated week (p > 0.15)Number of household members aged 7 17 (p > 0.15)Regular transport of children to school or kindergarten (yes = 1, no = 0) (p > 0.15)Pensioner (yes = 1, no = 0) (p > 0.15)Student/pupil (yes = 1, no = 0) (p > 0.15)Has moved to the present dwelling less than ve years ago (yes = 1, no = 0) (p > 0.15)Constant 135.91 49.85 0.0065 e number of days a person is present at the workplace or place of education is directly related to the number of weekly trips. eweekly number of working hours was tried as an alternative control variable, but this variable, too, showed a statistically non-signicanteect. Working hours were slightly negatively correlated with the distance from the dwelling to the city center, and including workinghours among the control variables therefore yielded a slightly stronger eect of residential location relative to the city center on the travelingdistance by car.


6040200

200

150

100

50

0

Control for sociodemographics, residential preferences,car ownership and car attitudesControl for sociodemographics and residential preferences

Figure 7: Average expected travel distances by car (km) over the veweekdays for each of the 29 investigated areas. e blackdots are based on the actual values of each urban-structuralvariable in the regression model, with socioeconomic vari-ables, demographic variables, and residential preferenceskept constant at mean values, cf. Table 4. (N = 1564 re-spondents, p = 0.0000.) e blue triangles are based on aregression analysis that includes, in addition to the variablesin Table 5, car ownership and attitudes toward car driving(N = 1476, p = 0.0000).

ably lower expected traveling distances by car than the otherperipheral residential areas. is reects the fact that boththese areas are located near second-order centers (the townsof Hillerd and Kge) and are also fairly close to urban railstations. Conversely, expected car usage in one residential arealocated about 20 km from the city center is clearly higher thanin the other residential areas located at similar distances fromthe city center. is reects the fact that the area in questionhas a particularly low local-area density and is located far fromthe closest second-order center and the closest urban rail sta-tion.

8 Concluding remarks

Our study shows that metropolitan-scale urban-structuralvariables generally exert stronger inuences thanneighborhood-scale built-environment characteristicson traveling distances by car during weekdays. In particular,the location of the residence relative to the main city center ofthe metropolitan region shows a strong eect on the amountof car travel. We also nd that the amount of weekdaycar travel is aected by the location of the dwelling relativeto the closest second-order center and to the closest urbanrail station, as well as the density of population and jobswithin the local area (a two-square-kilometer zone aroundthe residential area). Compared to these four variables,the eects of local-scale urban characteristics are generallyweaker. For example, when we control for the location ofthe dwelling relative to the city center, density measured atthe level of the narrowly demarcated residential area is not asstrongly correlated with traveling distances by car as densitymeasured within a larger geographical area. Similarly, thedistance from the dwelling to the closest facility within acategory is generally less important to traveling distancesthan proximity to concentrations of facilities. Since thehighest concentrations of service facilities and workplacesare found in the central and inner parts of the metropolitanarea, inner-city residents generally have better possibilitiesof nding suitable jobs, shopping opportunities, and leisurefacilities without having to travel long distances.

e nding that metropolitan-scale built-environmentcharacteristics exert a stronger inuence on travel behav-ior than neighborhood-scale characteristics is not limited toweekday car travel. Similar results have been obtained for totaltraveling distance onweekdays andweekends, and for the pro-portion of total distance traveled by car (Nss 2006b; Nssand Jensen 2005).

e obvious interpretation of these results is that the fourhigher-level urban-structural variables inuence travel behav-

An examination of 485 respondents who had moved from one resi-dence within the Copenhagen Metropolitan Area to another during theprevious ve years gives additional support to this claim. ese respondentswere askedwhether they, according to their own judgment, had experienceda change in their amountof travel due to themove. ephrasingof theques-tion was: If you have moved has moving from your latest to your presentresidence caused any changes in your amount of travel? e answer alter-nativeswere: a) Yes, moving has had the consequence that I now travelmoreb) Yes, moving has had the consequence that I now travel less c)No, movinghas not led to any changes in my amount of travel worth mentioning. eanswers to these questions show a clear tendency toward increased travelwhenmoving outward (Wald= 33.259, p = 0.0000) and decreasing whenmoving closer to the city center (Wald= 22.147, p = 0.0000).

.

ior through the accessibility of various types of facilities. Be-cause the variables measuring accessibility to dierent facili-ties only capture the travel purposes associated with the facil-ity categories in the respective indices, their eects are weakerthan the eects of the variables representing the location of theresidence relative to themain centers of themetropolitan area.Accessibility indices that include a greater number of facilitycategories generally exhibit a stronger relationship with travelbehavior. us, traveling distances by car are more stronglycorrelated with the index for availability of shopping oppor-tunities near the dwelling than with the index for distance tothe closest grocery store, and stronger with the combined in-dex for availability of service facilities near the dwelling thanwith the index for shopping opportunities.

Arguably, an equal or even higher statistical power of de-termination (adjusted R2) might have been obtained by re-placing the variables measuring distances to various types ofcenters with measures of accessibility at local, district, andmetropolitan scales Bhat and Guo (2007); Krizek (2003).However, as guidance for urban planning, it is probably moreinteresting to know how the location of the dwelling relativeto various types of centers aects travel behavior than to knowthe relationship between travel behavior and, for example, themean opportunity distance.

e moderate eect of local-area density on traveling dis-tances by car should not lead us to believe that neighborhood-scale density is unimportant to travel. Apart from inuencingthe provision of local services and public transport, local areadensities add up to the overall density of the city. e higherthe population density of the city as a whole, the lower willbe the average distance between the residences and the down-town area. In this way, local area densities indirectly inuencethe urban-structural variable that, according to our studies, ex-erts the strongest inuence on the travel behavior of individu-als and households, namely the location of the residence rela-tive to the city center.

Interestingly, any relationship between the local-level streetstructure and traveling distance by car disappears with the in-troduction of statistical control for the location of the resi-dence relative to downtown Copenhagen. is gives rise to asuspicion that the corresponding relationship seen in researchcarried out in the United States might reect the location ofthe residential areas rather than the shape of the local streetnetwork. In most of the American studies that have attacheda great importance to the shape of the local street pattern,control for the location of the residential area relative to thehigher-level center structure seems to be missing.

e results of this study are in line with the ndings of anumber of studies in other cities in Europe and Asia, as notedin the introductory section. e Copenhagen study nd-ings are also in accordance with evidence from some Ameri-can studies, such as (Ewing and Cervero 2001) and (Zegras2010), both of which found regional accessibility to be moreimportant than local built-environment characteristics to thenumber of vehicle miles traveled. A clear dierence in theamount of car travel between suburban/rural residents andresidents living close to the Central Business Districtalsoaer control for socioeconomic and attitudinal factorswasalso found in a study by (Zhou and Kockelman 2008).

e lesson for spatial planners aiming to facilitate more en-vironmentally friendly travel patterns in city regions is that ur-ban containment ismore conducive to this end than the devel-opment of new suburbanneotraditional housing areas. In par-ticular, densication close to themain center of themetropoli-tan area contributes to a reduction in traveling distances andencourages the use of travel modes other than the private car.From the perspective of sustainable mobility, metropolitan-level centralization is thus more favorable than decentralizeddevelopment according to New Urbanist principles. Today,fortunately, many European and American cities have consid-erable opportunities for inner-city densication and regener-ation due to the strong deindustrialization processes that havebeen going on during recent decades in mostWestern cities.

References

Bhat, C. and J. Guo. 2007. A comprehensive analysis ofbuilt environment characteristics on household residen-tial choice and auto ownership levels. TransportationResearch Part B: Methodological, 41(5):506526. doi:10.1016/j.trb.2005.12.005.

Boarnet, M. and R. Crane. 2001. Travel by design: e inu-ence of urban form on travel. Oxford andNewYork: OxfordUniversity Press. ISBN 9780195123951.

Cao, X., P. Mokhtarian, and S. Handy. 2009. Examin-ing the impacts of residential self-selection on travel be-haviour: A focus on empirical ndings. Transport Reviews,29(3):359395. doi: 10.1080/01441640802539195.

Cervero, R. 1989. Jobs-housing balancing and regionalmobility. Journal of the American Planning Asso-ciation, 55(2):136150. ISSN 0194-4363. doi:10.1080/01944368908976014.

Cervero, R. 2003.ebuilt environment and travel: Evidencefrom the United States. European Journal of Transport andInastructure Research, 3(2):119137.


Chatman, D. G. 2005. How the Built Enironment InuencesNon-work Travel: eoretical and Empirical Essays. Ph.D.thesis, University of California, Los Angeles, Los Angeles.

Christensen, L. 1996. Location-dependence of transportationtrends: e importance of city class. Stockholm, Sweden:Danish National Environmental Research.

Ewing, R. and R. Cervero. 2001. Travel and the built en-vironment: A synthesis. Transportation Research Record,1780(1):87114. ISSN 0361-1981. doi: 10.3141/1780-10.

Fourchier, V. 1998. Urban density and mobility in ile-defrance region. In Proceedings of the Eighth Conferenceon Urban and Regional Research, pp. 285300. Madrid:UN/ECE-HPB andMinisterio de Fomento.

Fox, M. 1995. Transport planning and the human activityapproach. Journal of Transport Geography, 3(2):105116.ISSN 09666923. doi: 10.1016/0966-6923(95)00003-L.

Frank, L. D. and G. Pivo. 1994. Relationships between landuse and travel behavior in the Puget Sound Region. FinalSummary Report WA-RD 351.1, Washington State Trans-portation Center, Seattle.

Handy, S., X. Cao, and P. Mokhtarian. 2005. Corre-lation or causality between the built environment andtravel behavior? evidence from Northern California.Transportation Research Part D: Transport and Eni-ronment, 10(6):427444. ISSN 13619209. doi:10.1016/j.trd.2005.05.002.

Handy, S. and K. Clion. 2001. Local shopping as astrategy for reducing automobile travel. Transportation,28:317346.

Handy, S., K. Clion, and J. Fisher. 1998. e eectivenessof land use policies as a strategy for reducing automobiledependence: A study of Austin neighborhoods. ResearchReport UCD-ITS-RR-98-14, Institute of TransportationStudies, University of California, Davis, Davis. URL http://pubs.its.ucdavis.edu/publication_detail.php?id=510.

Hgerstrand, T. 1970. Urbaniseringen af Sverige en ge-ogrask samhllsanalys. (e urbanization of Sweden ageographical analysis of society). SOU, 14:Appendix 4.

Jones, P. 1990. Developments in Dynamic and Activity-basedApproaches to Travel Analysis. Aldershot; Brookeld: Ave-bury. ISBN 9780566070235.

Krizek, K. 2003. Residential relocation and changes in urbantravel: Does neighborhood-scale urban formmatter? Jour-nal of the American Planning Association, 69(3):265281.doi: 10.1080/01944360308978019.

Maddison, D., C. for Social, and E. R. on the Global Envi-ronment. 1996. e true costs of road transport. London:

Earthscan. ISBN 9781853832680.Mogridge, M. 1985. Transport, land use and energyinteraction. Urban Studies, 22(6):481492. doi:10.1080/00420988520080851.

Newman, P. and J. Kenworthy. 1989. Cities and automobiledependence: A sourcebook. Aldershot; Brookeld: Gower.ISBN 9780566070402.

Nielsen, T. S. 2002. Boliglokalisering og transport i Aalborg.(Residential location and transport in Aalborg). PhD disser-tation, Aalborg University, Aalborg, Denmark.

Nielsen, T. S. and H. Hovgesen. 2006. Developments in therelationship between large cities the suburban zone andthe rural/urban hinterland. InCities in City Regions. Euro-pean Urban Research Association (EURA).

Nowlan, D. and G. Stewart. 1991. Downtown popu-lation growth and commuting trips: Recent experiencein Toronto. Journal of the American Planning As-sociation, 57(2):165182. ISSN 0194-4363. doi:10.1080/01944369108975485.

Nss, P. 2004. Prediction, regressions and critical real-ism. Journal of Critical Realism, 3(1):133164. doi:10.1163/1572513041172713.

Nss, P. 2005. Residential location aects travel behavior-but how and why? the case of Copenhagen metropoli-tan area. Progress in Planning, 63(2):167257. doi:10.1016/j.progress.2004.07.004.

Nss, P. 2006a. Accessibility, activity participation and loca-tionof activities: Exploring the links between residential lo-cation and travel behaviour.Urban Studies, 43(3):627652.ISSN 0042-0980. doi: 10.1080/00420980500534677.

Nss, P. 2006b. Urban Structure Matters: Residential Lo-cation, Car Dependence and Travel Behaviour. New York:Routledge, 1. ed. edition. ISBN 9780415375740.

Nss, P. 2009a. Residential location, travel behaviour,and energy use: Hangzhou metropolitan area com-pared to Copenhagen. Indoor and Built Enironment,18(5):382395. doi: 10.1177/1420326X09346215.

Nss, P. 2009b. Residential self-selection and appropri-ate control variables in land use: Travel studies. Trans-port Reviews, 29(3):293324. ISSN 0144-1647. doi:10.1080/01441640802710812.

Nss, P. 2010. Residential location, travel, and energy use:e case of the Hangzhou Metropolitan Area. Journal ofTransport and Land Use, 3(3):2759. ISSN 1938-7849.doi: 10.5198/jtlu.v3i3.98.

Nss, P. and O. Jensen. 2004. Urban structure matters,even in a small town. Journal of Enironmental Planningand Management, 47(1):3557. ISSN 0964-0568. doi:

.

10.1080/0964056042000189790.Nss, P. and O. Jensen. 2005. Bilringene og cykelnavet:boliglokalisering, bilangighed og transportadfrd i Ho-edstadsomrdet. Aarhus: Universitetsforlag. ISBN9788773077283.

Nss, P., P. Re, and S. Larsen. 1995. Travelling distances,modal split and transportation energy in thirty residen-tial areas in Oslo. Journal of Enironmental Planning andManagement, 38(3):349370. ISSN 0964-0568. doi:10.1080/09640569512913.

Rajamani, J., C. Bhat, S. Handy, G. Knaap, and Y. Song.2003. Assessing impact of urban form measures on non-work trip mode choice aer controlling for demographicand level-of-service eects. Transportation Research Record,1831(1):158165. doi: 10.3141/1831-18.

Roundtable, C. P. 2008. Deconstructing Jobs-Housing bal-ance. Technical report, California Planning Roundtable.

Schwanen, T., F. Dieleman, and M. Dijst. 2001. Travel be-haviour in dutch monocentric and policentric urban sys-tems. Journal of Transport Geography, 9(3):173186. ISSN09666923. doi: 10.1016/S0966-6923(01)00009-6.

Sieverts, T. 1999. Zwischenstadt: Zwischen Ort Und Welt,Raum und Zeit, Stadt und Land. (e between city be-tween place and world, space and time, city and countryside).Wiesbaden: Vieweg. ISBN 9783764363932.

Stead,D. and S.Marshall. 2001.e relationships betweenur-ban form and travel patterns: An international review andevaluation. European Journal of Transport Inastructure Re-search, 1(2):113141.

Steg, L., C. Vlek, and G. Slotegraaf. 2001. Instrumental-reasoned and symbolic-aective motives for using a mo-tor car. Transportation Research Part F: Trac Psychol-ogy and Behaviour, 4(3):151169. doi: 10.1016/S1369-8478(01)00020-1.

Urry, J. 2000. Sociology Beyond Societies: Mobilities for theTwenty-First Century. New York: Routledge. ISBN9780415190893.

Vibe, N. 1993. Norske reisevaner. dokumentasjonsrapportfor den landsomfattende reisevaneunderskelsen 1991-92. Technical Report 183, Transportkonomisk institutt,Oslo.

Vilhelmson, B. 1999. Daily mobility and the use oftime for dierent activities. the case of Sweden.GeoJournal, 48:177185. ISSN 0343-2521. doi:10.1023/A:1007075524340.

Weitz, J. 2003. Jobs-housing balance. Technical ReportPlanningAdvisory ServiceReportNo. 516, AmericanPlan-ning Association, Chicago. URL www.planning.org/pas/

reports/subscribers/pdf/PAS516.pdf.Zegras, C. 20

170-1354-1-PB

Documents

Transcript of 170-1354-1-PB